As a trader, you probably noticed understand that markets change from time to time. For example, during the summer we often see that the markets are quieter and during the fall we usually get some more volatility. It’s important for you to adapt to these changes A good trading strategy can withstand these changes in the markets. But how do you know if your trading strategy is good and robust enough? That’s exactly where a Monte Carlo simulation can help. Apart from using your historical data to learn more about your trading performance, you can perform a Monte Carlo Simulation formula to get an idea of how your strategy might perform in the future. In this article guide, I’m going to try to explain the Monte Carlo Simulation for you so in the end, you know how you can use it to analyze and improve your trading results.
A Monte Carlo Simulation is a simulation that applies random changes to historical trades and then calculates a new equity curve. This new equity curve is then used to verify if the trading strategy is robust enough to withstand the random changes.
This technique has been widely used by other professionals in various fields such as physical science, energy, biology, artificial intelligence, project management, transportation, research and development, insurance, environment, oil, and gas among others. It’s widely used in many disciplines because it’s able to give a wide range of possible outcomes and the probability of them occurring in a given model based on different parameters that aid in key decision making. Although the Monte Carlo Simulation gives a probability of outcomes happening and it’s not deterministic, it serves as a good estimation tool that helps decision-makers get a possible approximation of future reality.
In trading, we use Monte Carlo simulations to analyze how our trading strategy would have performed if for example, we had slightly worse entries or exits, or if we missed a few good trades.
The Monte Carlo Simulation technique was first used by Stanislaw Ulam, who was a mathematician working on the Manhattan atom bomb project. After World War II, he was recovering from brain surgery when he decided to keep himself entertained by playing unlimited games of solitaire. It was at this point where he became curious and he wanted to know whether it’s possible to predict the outcomes of each of the games of solitaire. He wanted to observe their distributions so that he can determine his probability of winning these games. He later shared the idea with John Von Neumann who collaborated with him to invent the Monte Carlo Simulation technique. Monte Carlo Simulation was named after Monte Carlo, which is Monaco’s most renowned town spot for gambling. Just like in gambling which is associated with chances and random outcomes, the Monte Carlo Simulation technique uses the same concept to predict possible outcomes and the probability of them occurring in a given model.
Monte Carlo Simulation works by taking all the trades you made in the past and applying small random changes to these trades.
Some example of what might be changed:
After applying the random changes to all your trades, it will then use these changed trades to calculate a new equity curve. This new equity curve already gives you a clue about how good your trading strategy is able to survive these random changes. However, a single simulation does not give a really good accurate result. We normally perform 100s of these Monte Carlo simulations. In each simulation, it will make random changes to your trades like modifying your entry/exit a bit or changing the order of the trades. Each simulation gives a new potential equity curve and by looking at 100s of these simulations, we get a pretty accurate idea of how robust your trading strategy is and if it can survive small market changes, and what your drawdown or risk to ruin could be like.
Monte Carlo Simulation uses the probability distributions to describe uncertainty in variables when performing risk analysis in a given situation so as to aid in the right decision making. When probability distributions are used in a given model, the variables used have higher chances of resulting in different probability outcomes and that’s where the Monte Carlo Simulation technique comes in to provide data of possible future outcomes.
There are different types of probability distributions that are used in Monte Carlo Simulation calculations. Some of the common probability distributions are;
Here just as the name suggests, all the values have an equal chance of occurring. The user, in this case, will only be interested in determining the minimum and the maximum values. A good example where the uniform probability distribution is used is in the calculation of production costs in companies to predict the future sales revenues of a new product or service.
b) Normal or Bell Curve
Under the normal or “bell curve” probability distribution, the user's aim is to define the mean and standard deviation so as to describe the variation. Under the normal curve, all the values which are in the middle and near the mean have a high probability of occurrence. A normal/bell curve is presented as symmetric and it describes a wide variety of natural occurrences and features. A normal probability distribution can be used to describe numerous variables outcome such as changes in inflation rates as well as energy price levels.
The Lognormal probability distribution is different from the normal distribution in that the values are positively skewed as opposed to being symmetric. It’s mostly used to analyze values that don’t go below zero digits but they have infinite positive potential. A lognormal probability distribution is used to describe variables such as oil reserves, stock market prices, and real estate property prices
In triangular probability distribution, the user's aim is to define the minimum, most likely, and the maximum values. Under this scenario, it’s the values that fall under the most likely category which have the highest potential of occurrence. Different product inventory levels and past sales history per unit of time are some of the variables that can be expressed via a triangular distribution.
Under PERT, the user's aim is to determine the minimum, most likely, and maximum values which is the same as the triangular probability distribution. However, the two are different in the sense that in PERT distribution, there are more chances of values between the most likely and the extremes are more likely to occur. In triangular distribution, the values on the extremes are not given many considerations. A good example where PERT probability distribution is normally used is in the analysis of task duration in a certain project management model.
When using a discrete probability distribution, a user is interested in defining specific outcomes that have the potential of occurring and the possibility of each of them. For instance, a user may want to predict the outcomes of a lawsuit that have different specific possible outcomes such as a 30% chance of positive ruling, a 10% negative ruling, a 15% chance of mistrial, or a 45% chance of settlement.
In order to predict and quantify future risks and possible outcomes in trading, the Monte Carlo Simulation technique samples various variables randomly from the input probability distribution model. Every set of samples taken is known as an Iteration. All the resulting outcomes from this set of sample is carefully recorded and is later compared to other outcomes from repeated simulations (hundreds or thousands) of different iteration. The trader then interprets the data from the Monte Carlo Simulation calculations to understand the risk probability in the trading strategy so as to make an informed decision. Performing Monte Carlo Simulation calculations will not only tell you what could happen in the future but also the probability of it occurring.
Monte Carlo simulations can help you to determine if your trading strategy is robust enough to survive small changes in the financial markets and your trading. Analyzing its results will help you to further improve your trading performance.
The first advantage of using Monte Carlo Simulations is that you get a good idea of the maximum drawdown you can expect to encounter. Even if you only experienced a drawdown of let’s say 8% so far, A Monte Carlo simulation could tell you that it's very likely that you might encounter a 20% drawdown somewhere in the future. By knowing this in advance you are mentally prepared for it, which makes it easier to just keep following your trading plan when it happens. You can also try to optimize your trading strategy further to get a lower maximum expected drawdown.
The second advantage of a Monte Carlo simulation is that it will tell you your risk of ruin. I’ve talked about the risk of ruin over here. In my opinion, it’s one of the most important trading metrics every trader should know. Although you can use Kaufmanns or Ralph Vince’s formula to calculate the risk of ruin, a Monte Carlo Simulation will give you an even more accurate result since it's based on 100s of simulations and takes into account all kind of small changes which will happen during your trading.
As discussed in the trader's death spiral, most traders lose confidence in their trading strategy after a few losses. They will go out and find a new trading strategy and trade that for a while. Then the inevitable losing streak happens again and they lose confidence again and so they keep going in circles. One of the advantages of a Monte Carlo Simulation is that it will tell you what your winning- and losing streak might be. By knowing this upfront you won’t freak out and lose your confidence in your strategy when this actually happens. Just remember that losing is part of trading and a winning streak is just as normal as a losing streak. And if you know that it’s normal to have for example 6 losers in a row, then that makes it much easier to stick to your trading plan and ride it out.
Just like any other formula, there are a few downsides of Monte Carlo Simulation analysis when used in trading. One disadvantage is that it requires a trader to have a profitable trading strategy. If a trader keeps on changing their trading strategy, then the historical trade data used will lose their fundamental meaning with time when you continue trading. For you to get a valid Monte Carlo Simulation analysis, you need to have consistent historical results that are backed by a single robust strategy or method for the whole set of trading.
Monte Carlo Simulation analysis does not incorporate big fundamental changes in the markets. For example, the financial markets might change drastically due to disasters, wars, COVID 19, etc. When things like this happen and we see huge changes in the market then remember that carrying out a Monte Carlo Simulation analysis based on historical data might not give a valid future result. This is because the market has fundamentally changed over time.
When Monte Carlo Simulation strategy is used in a trading system, it involves taking random simulated trade series in order to understand whether the trading strategy is good enough and robust to withstand small changes in the market. When you have an idea of what the future trading or market may react, you are able to make an informed decision based on quantified market data.
Monte Carlo Simulation analysis can help you test the robustness of your trading strategy in order to understand a number of key trading performance metrics such as; Risk of Ruin, Annual rate of returns, Drawdown ratios and Maximum and Median drawdowns. When a trader has such important information at their disposal before they start live trading, they will be in a better position to make well-informed decisions when it comes to their trading strategy choice, capital allocation as well as position sizing.
There are a number of methods that are used to perform Monte Carlo simulation analysis in a trading system but they differ in their implementation components. Some of the most commonly used ones are;
This is the most straightforward and widely used Monte Carlo simulation method in trading systems. It involves taking the historical trade results and reorganizing their order. The simulation is done 1000 times to get our “new” 1,000 equity curves which gives us more information regarding the potential risks that we may experience when we decide to trade this particular strategy. When using this simulation method, our fundamental assumption should be that our trade results should always remain the same but they should not appear in the same order like the past ones.
Figure 1: Monte Carlo Simulation- Original/Resample
Once you get the simulation results, you can now use this information to calculate a number of key trading performance metrics such as the maximum or average drawdown of the “new” 1,000 equity curves. In most cases, when you calculate the average drawdown, you will realize that is always higher than the backtest’s drawdown. Rather than relying on the original backtest’s drawdowns, the data gathered from the average drawdown analysis can help you in making the right decision when it comes to trading system allocations and sizing.
Such simulation analysis can help a retail trader have better expectations when implementing a trading strategy. For instance, if your trading system had a backtest drawdown of about 10% and you sized your trades based on this data, if your live trading produces a drawdown of about 15%, you will be forced to turn off your new trading strategy. However, if you could have used the original and resample Monte Carlo Simulation analysis method, you could have been aware of the probability of a drawdown of this level occurring, prompting you to modify your trade sizing.
Another advantage of this simulation method is that it gives you great insights on when to expect profitability. For instance, if you are just trading casually, you can make a general guess that your trade strategy is no longer profitable after making 30 consecutive trades. If this scenario happens, you might be tempted to turn off the trading strategy because you can’t predict future results if you continue trading with this new strategy. On the other hand, if you had taken some time and analyzed your new trading strategy using Monte Carlo simulation, you could have understood that there was a likelihood that you will not make any profits until you reach trade 40 and above.
The randomized Monte Carlo method is mostly used to find out if there are any overfitting in the trading strategy creation process. This method is also known as the bootstrapping method. It performs a random sampling by re-trading each strategy entry signal from the backtest and replacing it with a random but appropriate exit for every signal. The process is then repeated 1,000 times to give us our “new” 1,000 equity curves for another in-depth performance analysis.
Under the randomized Monte Carlo simulation analysis method, the assumption is that if our trading signal is strong enough, we expect it to generate profits irrespective of the exit (s) chosen. When we use random exits, we expect the returns will not be as smooth as it is the case with the original exits but at least we expect the signal to still have some sort of profits at the end of simulation. If in any case the trading system does not maintain some level of profitability, there is a higher chance that the original entry signal that was used had an overfit to the historical data and you should consider dropping-off the trading strategy before you can even risk your investments.
There are two key trading properties that can be randomized when carrying out a Monte Carlo simulation analysis. The two trading properties are;
1) Change of Trades Order- When changing the trade order, there are two likely possibilities. The first one is where we randomly shuffle the original order of the trades. The other random resampling variation involves more than just shuffling the trades. It goes further and randomly picks the overall number of trades from all the trades available.
The second one is different from the first one in the sense that the list of trades might not necessarily be the same. Therefore, it has the potential of picking one trade multiple times and not picking other trades at all.
2) Skipping some Trades- Under this consideration, we deliberately make some trades to be randomly missed based on a given probability. The test helps a trader get an idea on how the equity curve would appear if some of the trades were randomly skipped. In a real trading scenario, you can miss a trade if there is a system failure or when you stop trading for a while.
You follow the below steps when carrying out a bootstrapping test on your new trading system in order to analyze your new trading strategy performance under different market scenarios.
Here is a data chart sample of randomized Monte Carlo simulation in trading;
Figure 2: Monte Carlo Simulation- Randomized
Apart from helping us quantify the risks that our new trading strategy might experience, the Monte Carlo Simulation exercise can also help us to size our new trading strategy. This is done by carrying out the Monte Carlo Drawdown test that allows us to assign a certain percentage of confidence to what levels of potential risks that we may assume in our new trading strategy.
To carry out Monte Carlo Simulation Drawdown test, you follow the below steps;
• The user starts by specifying their starting capital amount
• The next step is to run the Original or Resample Monte Carlo Simulation
• Create a record for all the 1,000 maximum drawdowns as a percentage of the starting capital
• Plot for another 1,000 maximum percentage drawdowns as a frequency distribution
• On the Y-axis, follow the cumulative distribution line all the way up to the 95% mark
• On the X-axis, find the corresponding X-axis drawdown percentage value which matches the 95% value on the Y-axis.
From this example, we can conclude that 95% of all the Monte Carlo simulations are below 20% drawdown as indicated on the X-axis. Therefore, we are now confident that we only have a 5% chance of experiencing a drawdown which is greater than 20%. As a trader, if for instance, you are not comfortable with these numbers, you can always re-run the test with an increased starting capital and compare the results until you get the most appropriate drawdown amount which gives you the amount of confidence that you are looking for when trading.
The drawdown data allows a trader to have a properly sized trading strategy. When you size properly, you are able to make an informed decision based on the calculated risk, and from the results, you can be able to determine whether to continue with the new trading strategy or turn it off as soon as possible.
As we have already discussed above, the Monte Carlo simulation analysis helps traders to determine if their trading strategy is good and robust enough to withstand small changes in the market. The question of how many times should you run a Monte Carlo simulation always arises when a retail trader wants to analyze the performance of a new trading strategy. Ideally, the common practice is, you should repeat the simulation procedure multiple times so that you can make an informed decision based on the average results generated. This is in line with the theory of large numbers which requires multiple simulations to be done so as to arrive at the most accurate estimates.
In the context of Monte Carlo simulation analysis in trading, we base accuracy on the confidence levels in a particular trading strategy. If the confidence level is small, the value estimate generated from the trading is more robust. The most common and acceptable simulation confidence levels when testing the robustness of a new trading strategy range from 95% to 99%.
In most cases, a good value estimate in a trading system is generated when multiple Monte Carlo simulations are done in the range of 1,000-500,000 times. However, the time taken to generate all the estimated values depends on the complexity of the simulation algorithm used and simulator software type.
Monte Carlo simulation analysis, when used in trading, provides a trader with important information regarding their trading strategy on how strong it is when faced with future market changes and risks. There is a popular saying that states that, model results are as good as their historical data. If the historical data that was used in the analysis model was accurate and there are no significant changes in the market, then the Monte Carlo simulation analysis results will be more accurate. However, it’s important to note that the simulator does not factor in any unexpected real-life trading events which can happen when you are trading such as a market crash. Although the future results are based on some degree of probability and they may not be 100% accurate as of the reality, it helps retail traders to understand the trade-off that exists between risk and projected profits. A trader can, therefore, utilize this trading tool results to make the right investment decisions early enough such as knowing when to turn off any poor trading strategy or ceasing to live trade any underperforming trading system.
As a retail trader, before you even put any new trading strategy into real practice, you should consider running a Monte Carlo simulation so as to get a rough estimate of the expected drawdown and profits.
Apart from testing the robustness of your new trading strategy, the Monte Carlo simulation analysis also helps in determining the good expectancy levels and the number of simulations required. In most cases, your model should have 95% expectancy levels and you should run a minimum of 100 simulations in order to get more accurate results.
Monte Carlo simulations are used in a new trading strategy model to find out the probability of getting different outcomes in a complex trading process which cannot be easily predicted.
The Monte Carlo simulation analysis is a great tool that helps traders who have invested in different financial instruments to assess their trading system portfolio performance so as to make the right investment decision. Traders get to understand whether their trading strategy is good enough and robust to survive small changes in the market. The trading strategy results generated through multiple simulations give a trader an idea of how the trades are likely to perform in the future and the potential risks. With such information, traders are in better control of their trading and a trader can make a decision to cease trading using a poor strategy.
If you are a retail trader and you have never used automated trading systems before, you should consider investing in algo-trading tools like Monte Carlo simulation analysis so as to better your trading performance and improve results.
Do you want to know what a trading journal is, and why you should use one ? If you are serious about your trading, then you should treat it as a real business. This means you should track all your dat...
Read MoreIf you are looking to increase the amount of profit you are making from your trades and are failing to understand why you may not be, knowing the expectancy ratio of a trade before you make it can hel...
Read More